Advice on including or excluding interaction terms in RDD

[Karen Arulsamy]

Hi all,

I am studying the effect of a schooling reform on financial outcomes using a fuzzy regression discontinuity design (RDD). I have a methodological question on when to include or exclude interaction terms (treatment*distance from the cut-off)? I looked up papers and some do whereas some don't but they don't provide reasons for these choices. As I understand, the inclusion of interaction terms allow the slopes to differ on either side of the cut-off so it seems necessary.

Also, in my application, I can't obtain the F-Statistic when I include interaction terms (N = 1519) and since I need to use robust standard errors, I don't think its possible to calculate the F-statistic manually.

Thanks in advance.


Many thanks
Karen

[Jared Greathouse]

There's no RD design that I'm aware of that doesn't need interaction terms. But, you should be using statistical software to automate this, so you shouldn't need to make or generate anything.